Improving Siamese Networks for One-Shot Learning Using Kernel-Based Activation Functions

Jadon, Shruti; Srinivasan, Aditya

doi:10.1007/978-981-15-5619-7_25

Cited by 23 publications

(13 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Generative modeling is attractive for many reasons: i) Modelization of the latent space: Generative models express causal relations. ii) Generative models were used in semi-supervised learning settings, to improve classification [3,6,7,8,9].…”

Section: Related Workmentioning

confidence: 99%

Learning a Sparse Generative Non-Parametric Supervised Autoencoder

Barlaud¹,

Guyard²

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

Section: Related Workmentioning

confidence: 99%

Learning a Sparse Generative Non-Parametric Supervised Autoencoder

Barlaud¹,

Guyard²

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…Some maxpooling layers dropout layers were applied for avoiding the overfitting and reducing the network size. Although we used the Unet in this experiment, we note that the structure of the CNN can be flexible by introducing more one-shot deep learning-based networks architectures design tricks, works on this direction could be found in the literature ofKoch [14], Vinyals et al [22],Shaban et al [20], Chen et al [2], Jadon and Srinivasan [11]. An adam stochastic gradient optimizer with learning decay rate lr = lr/epochs number initialized to lr = 0.1 is used for training the network with the binary cross entropy loss function.…”

Section: Network Architecture and Training Detailsmentioning

confidence: 99%

One-Shot Learning for Landmarks Detection

Wang

Vandersteen

Raffaelli

et al. 2021

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Landmarks detection in a medical image is a mainstay for many clinical algorithms application. Learning-based landmarks detection is now a major successful methodology for many types of objects detection. However, learningbased approaches usually need a number of the annotated dataset for training the learning models. To tackle the lack of annotation issue, in this work, an automatic one-shot learning-based landmarks detection approach is proposed for identifying the landmarks in 3D volume images. A convolutional neural network-based iterative objects localization method in combine with a registration framework is applied for automatically target organ localization and landmarks matching. We investigated both qualitatively and quantitatively the performance of the proposed approach on clinical temporal bone CT volumes. The result shows that the proposed method is robust in convergence, effective in accuracy and reliable for clinical usage.

show abstract

“…The last layers of these networks are fed to a contrastive loss function layer, which calculates the similarity between the two inputs. The whole idea of using Siamese architecture [7] [5] is not to classify between classes but to learn to discriminate between inputs. So, it needed a differentiating form of loss function known as the contrastive…”

Section: B Siamese Networkmentioning

confidence: 99%

SSM-Net for Plants Disease Identification in Low Data Regime

Jadon¹

2020

2020 IEEE / ITU International Conference on Artificial Intelligence for Good (AI4G)

Self Cite

View full text Add to dashboard Cite

Plant disease detection is a necessary step in increasing agricultural production. Due to the difficulty of disease detection, farmers spray every form of pesticide on their crops to save them, causing harm to crop growth and food standards. Deep learning can help a lot in detecting such diseases. However, it is highly inconvenient to collect a large amount of data on all forms of disease of a specific plant species. In this paper, we propose a new metrics-based few-shot learning SSM net architecture, which consists of stacked siamese and matching network components to solve the problem of disease detection in low data regimes. We showcase that using the SSM net (stacked siamese matching) method, we were able to achieve better decision boundaries and accuracy of 94.3%, an increase of 5% from using the traditional transfer learning approach (VGG16 and Xception net) and 3% from using original matching networks. Furthermore, we were able to attain an F1 score of 0.90 using SSM Net, an improvement from 0.30 using transfer learning and 0.80 using original matching networks. The code is available on Github: https://github.com/shruti-jadon/Plants Disease Detection.

show abstract

Improving Siamese Networks for One-Shot Learning Using Kernel-Based Activation Functions

Cited by 23 publications

References 4 publications

Learning a Sparse Generative Non-Parametric Supervised Autoencoder

Learning a Sparse Generative Non-Parametric Supervised Autoencoder

One-Shot Learning for Landmarks Detection

SSM-Net for Plants Disease Identification in Low Data Regime

Contact Info

Product

Resources

About